{"id":"805c2549-0708-48df-ab1c-f88c696635b3","arxiv_id":"2411.11679","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A foundry-fabricated Si3N4 microring produces 3.7 dB of directly detected twin-beam squeezing, consistent with an overcoupling model and implying roughly 10.7 dB on-chip.","lead":"The authors report 3.7 dB of directly detected quantum squeezing from a foundry-made silicon nitride microring, with an inferred on-chip squeezing of about 10.7 dB. The result is a record for continuous-wave nanophotonic squeezing and indicates CMOS-compatible chips could generate useful squeezed light for quantum sensing.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claim that the 3.7 dB is free of excess-noise degradation rests on an assumption the paper itself contradicts for nearby devices; the loss-dilution check in Fig. 3b cannot distinguish excess optical noise from squeezed vacuum.","rationale":"The reader's weakest assumption correctly identifies excess classical noise rejection as the key fragility, and the paper's own text provides direct evidence that this assumption fails in nearby parameter regimes: higher-threshold rings and the most-overcoupled 450-GHz ring were excluded from the stable-squeezing claim because of excess noise. My analysis adds that the loss-dilution check in Fig. 3b cannot resolve this concern, because an excess-noise term added to the optical beams before the variable loss is attenuated by the same factor as the squeezed vacuum; both contributions approach shot noise as eta_path goes to 0. The fit of Eq. (1) with eta_path as a free parameter over three points with error bars of roughly 0.2 dB does not have enough statistical power to exclude a contamination that would shift the headline from 3.7 dB to, say, 3.3 dB. That said, the paper has genuinely independent support: the shot-noise calibration is linear over 24 dB of power, the dark noise is subtracted, and the measured 3.7 dB is consistent with the independently measured eta_path = 77% and eta_D = 79% for the blue point. The reported result is therefore plausible but not settled. A frequency-resolved noise measurement and a CMRR check at the operating point would directly test whether the residual classical noise is truly negligible, and would either validate the central claim or reduce it to a conditional upper bound. The reader's CONDITIONAL verdict is appropriate, so I do not recommend changing it.","tokens_in":12666,"tokens_out":14049,"duration_ms":143860,"concrete_test":"Re-measure the headline device (theta = 0.914, 3.7 dB point) while recording the intensity-difference noise spectrum from 1 to 50 MHz at fixed optical power, with the PID off and on. Also measure the balanced detector's CMRR at 5 MHz by injecting a sinusoidally modulated pump and recording the differential output. If the squeezing is flat across the band and the CMRR is at least 35 dB at 5 MHz, excess-noise contamination is unlikely; if the squeezing degrades below 10 MHz or tracks the EDFA RIN spectrum, the reported 3.7 dB is degraded by classical noise. An independent check: repeat the shot-noise calibration using tunable laser light at 1521 nm and 1587 nm (the twin-beam wavelengths) to verify that the wavelength dependence of eta_D and the grating does not shift the reference by more than 0.1 dB.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central result -- that the detected 3.7 dB represents the predicted quantum squeezing with no measurable degradation from classical excess noise -- requires that at the 5 MHz analysis frequency the EDFA RIN, pump intensity noise, and thermal fluctuations contribute negligibly to the intensity-difference signal after the ~35 dB CMRR. This is the least secure condition. Section III reports that devices with threshold power above 250 mW showed no squeezing commensurate with theta, and that the most-overcoupled 450-GHz device (theta = 0.93, the orange point used for the 11.5 dB on-chip inference) did not yield the stable squeezing trace of Fig. 2, 'likely attributable to the increased excess noise associated with higher pump power.' If the same excess-noise mechanism partially contaminates the three reported points, the directly detected value is an upper bound, and the 'no degradation' conclusion plus the on-chip 10.7-11.5 dB inference via Eq. (1) are overestimates. The loss-dilution test in Fig. 3b does not close this gap: optical excess noise added before the loss element is attenuated by exactly the same factor eta as the squeezed field, so the measured noise still approaches shot noise as eta goes to 0. Thus the distinguishing signature is absent. Only three operating points (theta = 0.875, 0.914, 0.93) support the fit, and the fit uses eta_path as a free parameter; this is too weak to certify the absence of a contamination that is explicitly seen in neighboring devices.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports the direct detection of 3.7 ± 0.2 dB of twin-beam intensity-difference squeezing from a foundry-fabricated Si3N4 microring operated as a Kerr optical parametric oscillator above threshold. The authors demonstrate stability of the squeezing over ~10 s, robustness to small pump detuning changes, and consistency with the prediction of Eq. (1) across three devices/resonances with overcoupling coefficients θ = 0.875, 0.914, and 0.93. A one-parameter fit to the θ-trend returns ηpath = 77%, matching an independent measurement, and a loss-dilution scan in Fig. 3b shows the detected squeezing approaching shot noise as added loss increases. The paper also reports an on-chip inferred squeezing level, quoted as 10.7 dB in the abstract and 11.5 dB in Section III, and attributes the low oscillation threshold of 450-GHz-FSR rings to reduced thermal/excess noise.","tokens_in":13030,"tokens_out":9183,"duration_ms":85673,"significance":"If the excess-noise-free interpretation is correct, this is a significant advance: it would be the largest squeezing directly detected from a microresonator, in a CMOS-compatible platform, with an inferred on-chip squeezing level above 10 dB. The paper's strengths include an external shot-noise calibration (Fig. 2b), an independent verification of the single fitted parameter ηpath, repeated measurements across devices and detunings, and an explicit threshold-power comparison between 450-GHz and 210-GHz FSR rings. The central limitation is that the 'no noticeable degradation from excess classical noise' claim is not fully established: the paper itself concedes that nearby devices and even the most overcoupled θ = 0.93 device show signs of excess-noise contamination, and the loss-dilution test in Fig. 3b cannot distinguish squeezed vacuum from excess optical noise added before the loss.","major_comments":[{"comment":"The abstract states that the squeezing is 'consistent with the overcoupling degree without noticeable degradation from excess classical noise,' but Section III explicitly reports that for the most overcoupled device (θ = 0.93, the orange point in Fig. 3a used for the 11.5 dB inference) 'the stable squeezing measurement in Fig. 2 was not seen,' with the stated explanation being 'increased excess noise associated with higher pump power.' This is an internal contradiction that directly affects the validity of the 11.5 dB on-chip claim. Please either provide a quantitative excess-noise characterization at that operating point that justifies including it, or exclude the θ = 0.93 point from the on-chip inference and restrict the headline on-chip claim to the θ = 0.914 point (10.7 dB).","section":"Section III, final paragraph; Abstract"},{"comment":"The loss-dilution test in Fig. 3b is presented as evidence that the measured squeezing is free of excess classical noise, but this test cannot distinguish squeezed vacuum from excess noise added before the loss element. If an excess-noise term N is present before the attenuation η, the detected variance is η(S + N) + (1 − η), which still approaches the shot-noise level as η → 0. Thus the observed linear degradation toward shot noise is equally consistent with a contaminated squeezed state. A control measurement that directly bounds the excess noise at 5 MHz is needed, for example measuring the intensity-difference noise with the pump below threshold, measuring the EDFA RIN at the detection frequency, or characterizing the balanced detector's CMRR in situ.","section":"Fig. 3b and surrounding text"},{"comment":"The inferred on-chip squeezing is quoted as 10.7 dB in the abstract and as 11.5 dB in Section III. These numbers correspond to different data points (θ = 0.914 versus θ = 0.93, respectively). The inconsistency is confusing and undermines the quantitative headline. Please harmonize the numbers and state explicitly which operating point supports each value.","section":"Abstract and Section III"},{"comment":"The θ-trend fit in Fig. 3a uses only three points spanning a narrow range (θ = 0.875 to 0.93) with ηpath as the single free parameter. While the agreement with the independently measured ηpath = 77% is a good cross-check, the fit cannot rule out a θ-dependent excess-noise component, especially because the highest-θ point is the one for which excess noise is explicitly acknowledged. Please report the fit residuals, the confidence interval for ηpath, and the result of excluding the θ = 0.93 point, so the reader can judge the robustness of the 'no degradation' conclusion.","section":"Fig. 3a and paragraph containing Eq. (1)"}],"minor_comments":[{"comment":"The caption labels the shot-noise trace both as 'Shot noise (yellow)' and later as 'Shot noise in black'; please clarify which curve corresponds to the directly measured trace and which is derived from the calibration in Fig. 2b.","section":"Fig. 2a caption"},{"comment":"The phrase 'SL ¿3 dB' appears to be a rendering artifact for 'SL > 3 dB'; please correct it.","section":"Introduction"},{"comment":"The caption notes that the spectral dependence of ηD and the grating efficiency are not accounted for; a short sensitivity estimate would help the reader understand how this uncertainty propagates to the fitted ηpath and the inferred on-chip squeezing.","section":"Fig. 3a caption"},{"comment":"The statement that 'the stable squeezing measurement in Fig. 2 was not seen' for the θ = 0.93 device should be quantified (e.g., fluctuation amplitude or Allan deviation) so that it is clear why the corresponding data point is nevertheless retained in Fig. 3a.","section":"Section III, θ = 0.93 discussion"},{"comment":"The denominator 1 + Ω²τc² is neglected because Ωτc ≪ 1; please state the maximum value of Ωτc for the three reported devices to make this approximation quantitatively transparent.","section":"Eq. (1)"}],"recommendation":"major_revision","confidential_remarks":"The direct 3.7 dB measurement appears to be a solid experimental result, and the external shot-noise calibration plus independent ηpath verification are commendable. The main risk is the overclaim in the abstract and Section III regarding the absence of excess-noise degradation and the corresponding 10.7/11.5 dB on-chip inference. The internal contradiction about the θ = 0.93 device must be resolved, and the loss-dilution test is not sufficient to certify the excess-noise-free interpretation. With additional control measurements or appropriately softened claims, the paper could become publishable; in its current form the quantitative headline is not fully supported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe direct 3.7 dB measurement is the real result here. It's the strongest directly detected continuous-wave squeezing from a microresonator, on a foundry Si3N4 platform, and the calibration is genuinely careful: the shot-noise reference is independently calibrated, the intentional loss-dilution scan behaves linearly, and the squeezing-versus-overcoupling trend fit uses a single free parameter (η_path) that lands on the independently measured value. That gives me reasonable confidence that the 3.7 dB number is real, and it is worth taking seriously.\n\nThe paper also does something honest that many would skip: it reports the devices that did not work. The 210-GHz rings with threshold power above 250 mW showed no squeezing, and the most-overcoupled 450-GHz device (θ = 0.93) did not produce a stable trace, attributed to excess noise. That openness is to the authors' credit, but it also undercuts the stronger claim in the abstract that the squeezing is 'robust across multiple devices' and that there is 'no noticeable degradation' from excess classical noise. If neighboring devices show exactly that degradation, the three points in Fig. 3a are too few to certify its absence for them. The stress-test note is right: the loss-dilution test in Fig. 3b cannot separate excess optical noise from squeezed vacuum, because both are attenuated by the same factor as the loss is increased. So the 'no degradation' conclusion is supported only by the consistency of the fit, not by a direct signature.\n\nThe inferred on-chip value also has an internal inconsistency the authors need to fix: the abstract and introduction quote 10.7 dB, while Section III gives 11.5 dB for the best point. That is a straightforward editorial error, but it matters because the on-chip claim is the headline for many readers. The direct 3.7 dB does not depend on that inference, but the 10.7–11.5 dB range does, and it inherits the excess-noise uncertainty.\n\nNone of this kills the paper. The direct measurement, the calibration chain, and the threshold-FSR insight are solid. The right fix is to either soften the inferred on-chip claim or add a direct test that separates classical excess noise from the squeezed field (for example, measuring at a different analysis frequency where the excess noise rolls off, or using a phase-sensitive detection that distinguishes the quadrature). A serious referee should engage with this, not desk-reject it.\n\nI'd bring it to the reading group and would cite the 3.7 dB direct result as a benchmark. Send it to peer review, expecting revision.","headline":"Direct 3.7 dB squeezing is real and well-calibrated; the inferred on-chip number and the excess-noise-free claim need more support.","tokens_in":13562,"tokens_out":2491,"would_cite":true,"duration_ms":23230,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper reports 3.7 ± 0.2 dB of directly detected quantum squeezing from a foundry-fabricated silicon nitride microring, the largest ever seen from a microresonator, with an inferred on-chip squeezing of 10.7 dB.","keywords":["squeezed light","silicon nitride","Kerr microresonator","twin-beam squeezing","four-wave mixing","optical parametric oscillator","quantum sensing","CMOS-compatible photonics"],"falsifier":"A direct, loss-independent measurement of the on-chip squeezing on the same device—for example, via heterodyne or homodyne tomography—that returned a value clearly below the inferred 10.7 dB, or a demonstration that detected squeezing improves less than the model predicts when detection efficiency is raised above the current 64%, would falsify the claim of loss-limited, excess-noise-free squeezing.","tokens_in":12490,"feed_emoji":"⚛️","tokens_out":9225,"duration_ms":54181,"temperature":0.7,"pith_summary":"This paper reports the highest level of quantum squeezing ever directly detected from a microresonator: 3.7 ± 0.2 dB of twin-beam intensity-difference squeezing, generated by a foundry-fabricated silicon nitride microring pumped above its four-wave-mixing threshold. The authors argue that the measured squeezing matches the analytic overcoupling model with no excess classical noise, allowing them to infer that the squeezing generated on-chip reaches 10.7 dB (up to 11.5 dB in the most overcoupled device). They also show that using small-radius rings with a 450 GHz free spectral range lowers the parametric oscillation threshold and stabilizes the squeezed state against thermally induced drifts. If correct, this establishes silicon nitride as a practical CMOS-compatible platform for on-chip quantum light sources, with direct relevance to quantum-enhanced sensing and information processing.","feed_headline":"3.7 dB squeezing from a foundry chip breaks microresonator record","feed_subtitle":"Directly detected on-chip squeezing hits 3.7 dB; 10.7 dB inferred at the source—a path to chip-scale quantum sensors.","key_machinery":"The load-bearing object is the above-threshold optical parametric oscillator model for four-wave mixing in a point-coupled microresonator, whose predicted squeezing level is $SL = 10 \\log_{10}(1 - \\eta_D \\eta_{\\mathrm{path}} \\theta)$, where $\\theta = 1 - Q_L/Q_i$ is the overcoupling coefficient. This formula, derived from the quantum noise theory of non-degenerate OPOs, converts a directly measured noise reduction into an on-chip squeezing level after accounting for detector inefficiency and optical loss, and it provides the quantitative test that the authors use to show their devices are free of excess noise. The second key mechanism is the dependence of the oscillation threshold on free spectral range, $P_{\\mathrm{th}} \\propto 1/\\mathrm{FSR}$, which motivates the use of 450 GHz FSR rings to reach low thresholds and stable operation.","core_discovery":"The central discovery is that a Kerr microresonator in the strongly overcoupled regime, operated just above parametric oscillation threshold, produces twin beams whose intensity-difference noise is 3.7 ± 0.2 dB below the shot noise level when detected directly, and that this value is consistent with the simple formula $SL = 10 \\log_{10}(1 - \\eta_D \\eta_{\\mathrm{path}} \\theta)$ given the known detector efficiency, path loss, and overcoupling coefficient. Across three device configurations with different overcoupling ($\\theta = 0.87, 0.91, 0.93$), the measured squeezing follows the predicted curve with a single fitted parameter—the path efficiency—which matches its independently measured value of 77%, indicating that excess classical noise does not measurably degrade the squeezing in these devices. The best device implies 11.5 dB of on-chip squeezing, and the paper reports that the detected squeezing is stable for minutes and robust to small pump detuning changes, thanks to active feedback and the lower threshold of larger-FSR rings.","pith_inferences":["Although the paper demonstrates squeezing only at a 5 MHz detection sideband, its finding that excess noise is absent in small rings suggests the same devices may also squeeze at lower frequencies where thermorefractive noise typically dominates; testing this could widen the usable bandwidth for quantum sensors.","The paper reports that the most overcoupled device ($\\theta = 0.93$) failed to squeeze stably, attributing this to excess noise at higher pump power; this hints at a design trade-off between overcoupling (which raises on-chip squeezing) and pump threshold (which raises noise), so an intermediate $\\theta$ might maximize usable squeezing.","Because the 10.7 dB on-chip value is inferred from the model rather than measured, an independent tomographic measurement of the same chip would be a natural validation step; the paper's framework gives a concrete prediction for what such a measurement should find."],"forward_implications":["Improving the total detection efficiency from its current 64% toward the 98% achieved in off-chip experiments would translate the inferred 10.7 dB on-chip squeezing into multi-dB directly detected squeezing from a chip, likely above 6 dB.","Because the squeezing matches the predictive model without excess-noise corrections, the same foundry process can be used to mass-produce squeezers that need no special noise mitigation, simplifying integration with CMOS electronics.","The combination of low threshold (via large FSR) and stable feedback locking makes these devices candidates for quantum-enhanced absorption spectroscopy and microscopy, where one twin beam probes a sample and the other serves as the reference.","Demonstrating that squeezing survives with 91% overcoupling suggests that further increasing overcoupling—if excess noise can be controlled—could push on-chip squeezing even higher, toward the 15 dB levels now available only off-chip."],"supporting_citations":[{"why":"Supplies the squeezing-level model Eq. (1) and the first on-chip Si3N4 squeezing (1.7 dB) that this work extends.","marker":"[31]"},{"why":"Recent Si3N4 OPO squeezing at 2.3 dB, the prior best continuous-wave nanophotonic result this work surpasses.","marker":"[33]"},{"why":"The record on-chip squeezer (4.9 dB pulsed) against which the continuous-wave performance is compared.","marker":"[44]"},{"why":"Provides the quantum noise theory of non-degenerate OPOs underlying Eq. (1).","marker":"[46]"},{"why":"Gives the quantum dynamics of Kerr frequency combs from which the squeezing formula is derived for this platform.","marker":"[47]"},{"why":"Provides the threshold-power scaling with FSR (Eq. 2) and demonstrates ultra-low-loss resonators enabling high Q.","marker":"[48]"},{"why":"Off-chip 15 dB squeezing used as the benchmark and for absolute quantum efficiency calibration.","marker":"[9]"},{"why":"A Si3N4 squeezing experiment that observed degradation from excess noise, providing the contrasting viewpoint this paper addresses.","marker":"[45]"}],"fun_headline_variants":["3.7 dB on-chip squeezing from a foundry-made SiN microring","Foundry chip sets squeezing record: 3.7 dB direct, 10.7 dB inferred","Quantum squeezing beyond 3.5 dB on a CMOS-compatible platform","Silicon nitride microring delivers 3.7 dB directly measured squeezing","Squeezed light from foundry chips reaches 3.7 dB, a new record"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim rests on the assumption that the residual classical noise from the erbium-doped fiber amplifier, thermal drift, and pump is fully rejected by the 7 nm filter and the balanced detector's common-mode rejection at 5 MHz, so that the measured intensity-difference noise reflects only the quantum state and the inferred on-chip squeezing of 10.7–11.5 dB is accurate.","fun_headline_variants_meta":{"raw":{"variants":["3.7 dB on-chip squeezing from a foundry-made SiN microring","Foundry chip sets squeezing record: 3.7 dB direct, 10.7 dB inferred","Quantum squeezing beyond 3.5 dB on a CMOS-compatible platform","Silicon nitride microring delivers 3.7 dB directly measured squeezing","Squeezed light from foundry chips reaches 3.7 dB, a new record"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000348,"raw_usage":{"total_tokens":1924,"prompt_tokens":983,"completion_tokens":941,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":599,"completion_tokens_details":{"reasoning_tokens":831}},"tokens_in":599,"tokens_out":941,"duration_ms":8828,"temperature":1.0,"reasoning_tokens":831,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T18:14:44.265205+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct, loss-independent measurement of the on-chip squeezing on the same device—for example, via heterodyne or homodyne tomography—that returned a value clearly below the inferred 10.7 dB, or a demonstration that detected squeezing improves less than the model predicts when detection efficiency is raised above the current 64%, would falsify the claim of loss-limited, excess-noise-free squeezing.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the squeezing-level model Eq. (1) and the first on-chip Si3N4 squeezing (1.7 dB) that this work extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Recent Si3N4 OPO squeezing at 2.3 dB, the prior best continuous-wave nanophotonic result this work surpasses."},{"cited_title":"Nehra, R","cited_arxiv_id":null,"evidence_quote":"The record on-chip squeezer (4.9 dB pulsed) against which the continuous-wave performance is compared."},{"cited_title":"Fabre, E","cited_arxiv_id":null,"evidence_quote":"Provides the quantum noise theory of non-degenerate OPOs underlying Eq. (1)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the quantum dynamics of Kerr frequency combs from which the squeezing formula is derived for this platform."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the threshold-power scaling with FSR (Eq. 2) and demonstrates ultra-low-loss resonators enabling high Q."},{"cited_title":"Vahlbruch, M","cited_arxiv_id":null,"evidence_quote":"Off-chip 15 dB squeezing used as the benchmark and for absolute quantum efficiency calibration."}],"review_version":1}